According to Gopi Prashanth of AI-startup Landing AI, one of the major challenges in the development of AI and machine learning is to teach machines to generalize based on small data samples in situations where relevant data is sparse. Currently, deep learning enables machines to learn from massive amounts of data, but Prashanth believes it is important that we re-engineer this approach so that it can work for sparse data samples as well.
Prashanth gives the example of quality inspection, an area where humans are good at determining if a product or part is faulty or functional: “Maybe one in 1,000 products are faulty, two at most; humans can take the two examples and generalize from them very well. But to teach a machine to use a few samples of data is a very hard technical problem to solve — it’s one of the key challenges we have to work on.” Prashanth believes that the solution will lie in teaching networks to generalize by studying “what features […] humans perceive and how do you train a model to do the same thing.”
Read more: AI must tackle the sparsity challenge, says Landing AI’s Gopi Prashanth